Guided Depth Map Super-Resolution Using Recumbent Y Network

Li, Tao; Dong, Xiucheng; Lin, Hongwei

doi:10.1109/access.2020.3007667

Cited by 11 publications

(7 citation statements)

References 60 publications

(131 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our experiments were conducted with the PyTorch framework [35] on a PC with an Intel(R) Core(TM) i7-8700K CPU @3.70GHz and an NVIDIA TITAN Xp GPU. We compared the proposed method with the baseline bicubic interpolation, the CNNs for the single image SR (SRCNN [36] and SAN [37]), and the recent color-guided depth map SR networks (MSG-Net [16], MFR-SR [18], DepthSR-Net [19], and RYNet [21]). We evaluated the performance of the conventional and proposed methods by conducting qualitative and quantitative comparisons for noise-free and noisy test data with various scaling factors (i.e., 2×, 4×, 8×, 16×).…”

Section: Resultsmentioning

confidence: 99%

“…Inspired by the residual U-Net architecture [24], Guo et al [19] proposed DepthSR-Net for depth map SR. DepthSR-Net learns the residual between the interpolated depth map and the ground truth HR depth map by using rich hierarchical features extracted from the network. Wen et al [20] introduced a coarse-to-fine CNN to approximate the ideal filter for depth map SR. Li et al [21] proposed a recumbent Y network (RYNet) for depth map SR. They built the network based on the residual channel attention blocks and utilized spatial attention based feature fusion blocks to suppress the texture copying and depth bleeding artifacts.…”

Section: B Color-guided Depth Map Srmentioning

confidence: 99%

“…Recently, owing to the success of deep learning, several color-guided depth map SR techniques [16]- [21] using convolutional neural networks (CNNs) have been proposed. One of the common characteristics of these techniques is to extract the features from the depth map and guidance image separately and then concatenate them such that the depth decoding or upsampling layers can inherit the HR details from the guidance features.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Depth Map Super-Resolution Using Guided Deformable Convolution

Kim

Baek

et al. 2021

IEEE Access

View full text Add to dashboard Cite

Depth maps acquired by low-cost sensors have low spatial resolution, which restricts their usefulness in many image processing and computer vision tasks. To increase the spatial resolution of the depth map, most state-of-the-art depth map super-resolution methods based on deep learning extract the features from a high-resolution guidance image and concatenate them with the features from the depth map. However, such simple concatenation can transfer unnecessary textures, known as texture copying artifacts, of the guidance image to the depth map. To address this problem, we propose a novel depth map superresolution method using guided deformable convolution. Unlike standard deformable convolution, guided deformable convolution obtains 2D kernel offsets of the depth features from the guidance features. Because the guidance features are not explicitly concatenated with the depth features but are used only to determine the kernel offsets for the depth features, the proposed method can significantly alleviate the texture copying artifacts in the resultant depth map. Experimental results show that the proposed method outperforms the state-of-the-art methods in terms of qualitative and quantitative evaluations.INDEX TERMS Convolutional neural network, depth map, super-resolution.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: B Color-guided Depth Map Srmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Depth Map Super-Resolution Using Guided Deformable Convolution

Kim

Baek

et al. 2021

IEEE Access

View full text Add to dashboard Cite

show abstract

“…For instance, a study by Hazrati et al [30] demonstrates the effectiveness of advanced frequency filtering techniques in cleaning audio signals in noisy environments. Similarly, research by Li et al [31] shows that frequency filtering strategies used in speech recognition systems contribute significantly to accurate speech recognition. These findings underscore the critical role of frequency filtering strategies in audio processing applications.…”

Section: Frequency Filteringmentioning

confidence: 87%

Investigation of the Effectiveness of Audio Processing and Filtering Strategies in Noisy Environments on Speech Recognition Performance

ÖZKURT

2024

Preprint

View full text Add to dashboard Cite

This study investigates the effects of audio processing and filtering strategies to enhance the performance of speech recognition systems in noisy environments. The focus is on the Short-Time Fourier Transform (STFT) operations applied to noisy audio files and noise reduction procedures. While STFT operations form the basis for detecting noise and analyzing the speech signal in the frequency domain, noise reduction steps involve threshold-based masking and convolution operations. The results obtained demonstrate a significant improvement in speech recognition accuracy in noisy environments through audio processing and filtering strategies. A detailed analysis of the graphs provides guidance for evaluating the effectiveness of noise reduction procedures and serves as a roadmap for future research. This study emphasizes the critical importance of audio processing and filtering strategies in improving the performance of speech recognition systems in noisy environments, laying a foundation for future studies.

show abstract

“…Finally, we use the attention mechanism CBAM [29] to fuse multiscale features from different branches. We apply this block in the last layer of the encoding network as in [30] to extract multiscale features.…”

Section: Encoding Networkmentioning

confidence: 99%

Inpainting larger missing regions via progressive guidance decoding network

Dong

Jiang

Hou

et al. 2023

Complex Intell. Syst.

Self Cite

View full text Add to dashboard Cite

For images corrupted for various reasons, the size of the corrupted area is often arbitrary and it has been a challenge to inpainting the larger missing areas. Though popular multistage networks ease the inpainting difficulty by repairing damaged image from coarse to fine, their common drawback is that the result of each stage is easily misguided by the wrong content generated in the previous stage. To address this problem, we propose a novel progressive guidance decoding network. First, multiple parallel decoding branches fill and refine the missing regions by top–down passing the reconstructed priors. This inpainting way of progressive guidance avoids adverse effects of inappropriate premises, since the decoding branches can learn what priors can be utilized. And convolution layers of decoder with different locations would pass down the different priors. The joint guidance of features and gradient priors helps the inpainting result contains the correct structure and rich details. The second fold of progressive guidance is achieved by our fusing strategy, combining ghost convolution and the designed cascaded efficient channel attention (CECA) to fuse and reweight the features from different branches. CECA explores the dependencies among adjant and non-adjant channels more effectively than popular ones. Finally, we merges the different-scale feature maps reconstructed by the last decoding branch and mapping them to the image space, which further improves the semantic plausibility of the restoration results. Extensive experiments verify the effectiveness of our method in both subjective and objective evaluation.

show abstract

Guided Depth Map Super-Resolution Using Recumbent Y Network

Cited by 11 publications

References 60 publications

Depth Map Super-Resolution Using Guided Deformable Convolution

Depth Map Super-Resolution Using Guided Deformable Convolution

Investigation of the Effectiveness of Audio Processing and Filtering Strategies in Noisy Environments on Speech Recognition Performance

Inpainting larger missing regions via progressive guidance decoding network

Contact Info

Product

Resources

About